Introduction and goals

For years Patricia Leopold (and others) of the Northern Institute of Applied Climate Science11 NIACS is “a collaborative effort among the Forest Service, universities, conservation organizations, and forest industry to provide information on managing forests for climate change adaptation and enhanced carbon sequestration. and I (and others) of The Nature Conservancy22 TNC has a mission to “Conserving the lands and waters on which all life depends”. have collaborated on tools and documents aimed at assisting managers with climate smart land management. We have hypothesized that we can add more value to land managers by coupling LANDFIRE33 LANDFIRE is an inter-agency program that “provides 20+ national geo-spatial layers (e.g. vegetation, fuel, disturbance, etc.), databases, and ecological models that are available to the public for the US and insular areas.” products and NIACS outputs. Lucky for us managers of the Deer Creek landscape are allowing us to use their area as a test case.

This document showcases some LANDFIRE-powered maps and charts for the Deer Creek landscape in West Virginia. In general LANDFIRE enables an assessment of past and present ecosystems and their disturbances plus the ability to make comparisons. LANDFIRE does not address climate change, or set “Desired Future Conditions”. 44 exploring potential impacts of climate change and how management actions may or may not lead to desired future conditions may be accomplished using SyncroSim and TNC’s modeling site for more information.

Due to the name and major funding sources LANDFIRE is often thought of as a fire dataset. While LANDFIRE does deliver many fire-focused datasets (e.g., the 40 Scott and Bergan Fire Behavior Fuel Models. Visit here) for more information], this assessment is focused on the vegetation datasets.

Additionally-this document is meant to be a conversation starter, and does not include the many possibilities brought to the table by NIACS for example.

A few characterisitics of LANDFIRE data and this report

LANDFIRE products are designed for use on large areas, typically comparing watersheds to watersheds for example. Often users ask “how small of an area can we use LANDFIRE data?” which is a tough question to answer as it varies based on data quality, heterogeneity of the vegetation (and underlying driving factors such as soils), and the particular dataset(s) being considered. Further, the purpose of data use is of particular importance. Using the data to explore trends is one thing, using LANDFIRE (or any data for that matter) to make place-specific decisions is another.

Schematic illustrating need for more review of data for smaller landscapes.

LANDFIRE data has been successfully used for mid-scale planning (~10k acres), but that was after considerable review by experts. To use data for smaller areas or for more detailed decision making requires more review per unit area (see graphic in sidebar).

LANDFIRE spatial data is delivered in raster format with 30m pixel size. Attribute tables can range from fairly simple (e.g., Existing Vegetation Cover), to rich with many fields (e.g., the Biophysical Settings data which includes historical fire regime information).

This report was created using free data and software. Data was downloaded from https://landfire.gov/getdata.php, all GIS processing and charts created in R55 R, and all maps created in Quantum GIS66 QGIS. Code available upon request.

The report itself was written as an R-Markdown77 Learn more about r-markdown documents here. document.

It is not necessary to use these tools! One could just as easily use ESRI and Microsoft products.

Which ecosystems were dominant prior to European Colonization?

With climate change, invasive exotic species, resource extraction and other modern ecosystem drivers it can be difficult to develop a baseline, or understanding of what was on the landscape prior to European colonization. LANDFIRE has a product suite under the Biophysical Settings (BpS)88 LANDFIRE defines Biophysical Settings as “the vegetation system that may have been dominant on the landscape prior to Euro-American settlement and is based on both the current biophysical environment and an approximation of the historical disturbance regime” label that includes three components. The BpS Descriptions99 find yours at here and accompanying state-and-transition models1010 download from this site. These documents and models describe how over 900 BpSs of the U.S. looked and worked historically. Each one includes a description of the major succession classes, and estimated amounts of each succession class based on natural disturbance regimes. Learn more about these products and their development by reading Blankenship et al. (2021).

Also included are the BpS spatial data1111 learn more here. This raster data includes a robust attribute table with information on historical fire regimes in addition to multiple classifications of the BpSs (e.g., there are lumped schemes).

Deer Creek dominated by Spruce-Fir and Dry-Mesic Oak Forests historically

Based on LANDFIRE BpS data three BpSs were present on over 60% of the landscape as shown in the chart and map below. Both graphics are not complete-there are more BpSs mapped than indicated on both. Lists were shortened for clarity, and to remove possible noise. A complete list and mapped amounts are available from the author.

Table of Biophysical Settings names, codes and acres

A zipped folder with Biophysical Settings descriptions is available here. Cross-reference the values in the “BPS_MODEL” column in the table above with the document name.

Also possible with the BpS products:

• quantification of historical disturbance amounts per year, including different types of fire and wind/weather/stress
• mapping of historical fire regimes
• delving into the state-and-transition models to explore and refine historical fire regimes to better match local knowledge, develop current condition models to document current disturbances and to modify to capture potential changes to disturbance regimes with climate change

In the last section of the document we will explore a couple ways to compare this historical data to current data.

---

What does the existing vegetation look like?

LANDFIRE delivers multiple datasets1212 visit the Vegetation page to get more information used to characterize current conditions. Below are charts and maps of Existing Vegetation Type (EVT), Cover (EVC) and Height (EVH).

The EVT data represents conditions as of ca 2016. This dataset is currently (January, 2022) to represent conditions as of ca2020. Release data TBD.

Deer Creek dominated by Dry-Mesic Oak and Montane Forests as of ca2016

Based on LANDFIRE EVT data there appears to be a loss of the Spruce-Fir forests. What is not clear from simply reviewing Bps and EVT data is an assessment of what converted to what. Additionally, colors were not matched between the BpS and EVT maps. Both graphics are not complete-there are more EVTs mapped than indicated on both. Lists were shortened for clarity, and to remove possible noise. A complete list and mapped amounts are available from the author.

Deer Creek dominated forests with relativity high canopy cover as of ca2016

As expected most of Deer Creek is mapped as the “forest” lifeform. It appears that most of the landscape is in a closed canopy condition.

Note: the data is delivered in 1% increments. In the graphics below the cover was categorized into 10% bins for viewing. I am happy to share raw attribute tables for further review.

Deer Creek dominated forests with most pixels mapped between 15-30M as of ca2016

As in the chart above most of the landscape is mapped as the “forest” lifeform. LANDFIRE mapped most of the pixels as having an average height between 15-30M with a range of 3-38M.

Note: the data is delivered in 1% increments. In the graphics below the cover was categorized into bins (different for each lifeform) for viewing. I am happy to share raw attribute tables for further review.

Additional analysis options with the Existing Vegetation data

With the Existing Vegetation products it is useful to do some additional GIS work that was not completed here. For example, we can combine datasets to explore height or cover patterns for a particular vegetation type as done here for an ecosystem in the Upper Cheat watershed:

Also beyond the scope of this “sample” would be combining LANDFIRE and non-LANDFIRE datasets. For example, some practitioners have combined Ecozone data from Steve Simon (and others) with LANDFIRE EVC and EVH as a way to map succession classes. In another use-case of this type I have combined EVT data with “Resilient Sites”1313 From the Center for Resilient Conservation Science data to explore how much of particular EVTs are in various levels mapped by this dataset for the Wayne National Forest1414 see draft report here.

---

Comparing historical and current vegetation

There are two main ways to compare historical with current vegetation conditions using LANDFIRE data:

1. Directly comparing the BpS and EVT data
2. Assessing over/under representation of succession classes1515 aka structural stages of each BpS. Learn more here

Below I present a couple ways to graphically represent these two types of comparisons.

Comparing the BpS and EVT datasets

In the BpS and EVT charts above it is possible to visually compare how amounts of different ecosystems have changed over time, but you cannot assess how a particular type has changed at a pixel by pixel level. To do so requires doing a “stack” or “combine” of the BpS and EVT datasets so each pixel gets both a BpS label and an EVT label. Once you have that in hand you can visualize the results in multiple ways. One possibility is a sankey diagram such as the one included below. Here I have greatly lumped vegetation types. It is possible to create a more detailed sankey charts without the lumping. Even with the lumping a few patterns emerge, for example a “conversion” of some of the Coniferous to Hardwoods.

Comparing the succession classes within particular BpSs

Each BpS has descriptions and mapping rules for up to 5 succession classes. These succession classes are largely definined by canopy cover and height, by some degree by composition. Using results from the models we get a “Reference” or expected percentage of each s-class. Using mapping we get the current percentages.

The following chart represents reference and current percentages of succession classes for 3 selected BpSs (ones that were fairly dominant historically).

From this chart a few patterns emerge:

• With the oak types there is an under-representation of the “open” succession classes, presumably due to fire suppression.
• In the spruce-fir BpS we see an under-representation of the “late-development” succession class
• The “reference” values may or may not represent a “desired future condition”

Next steps

As mentioned this document does not fold in any products from NIACS and in no way considers climate change. Hopefully we can find ways to merge this work with climate change science.

As noted it is possible to “combine” LANDFIRE with local datasets-this should be considered.

And finally, I have never seen Deer Creek! Review of all data needed before next steps are taken.

Appendix 1: Maps and charts for Monogahela NF Proclamation Area

LANDFIRE data is developed for large area use, typically areas greater than 100k acres. Use for areas smaller than that requires additional review. Due to the relatively small area of the proclamation area within the Deer Creek landscape the maps and charts are presented here instead of above.

Proclamation boundary was downloaded from: https://data.fs.usda.gov/geodata/edw/datasets.php, dataset named “Original Proclaimed National Forests and National Grasslands”. GIS processing and charts made in R, maps in QGIS. Code available upon request.